Ministry for National Economy

Hungarian Central Statistical Office

HCSO – Labour Market Department

Ministry for National Economy – 1051 Budapest, József nádor tér 4.

Hungarian Central Statistical Office – 1024 Budapest, Keleti Károly utca 5-7.

–

Hungary has conducted surveys similar to the Structure of Earnings Survey for more than 30 years. Since 1992 it has been an annual survey, with the reference month of May.

The requirements of the Structure of Earnings Survey were fulfilled by a few modifications and the development of the original survey in 2002 and since then all surveys have been conducted according to the EU requirements. The most important modifications included the following steps:

• The scope of the survey was extended to non-profit organizations;
• It was also extended to part-time workers;
• Some new variables were introduced including working time, overtime, length of service and paid vacation.

   In 2014 the National Labour Office accomplished the survey. The National Employment Office was the head office of the Public Employment Service in Hungary and – together with the Labour Centres– belonged to the Ministry for National Economy.

   

  Similarly to 2010, the survey in 2014 included micro-enterprises as well, including businesses with more than 2 employees. As previously, the response rate of micro enterprises was lower than the average response rate, however the Ministry decided to keep the micro enterprises in the survey every four years.

   

  The main uses of the survey are the following:

• It serves as a common database in the wage bargaining process in Hungary;
• It provides a sound basis for model computations to prepare minimum wage decisions and other wage agreements on the macro and medium levels;
• It helps prepare important decisions concerning the development of different wage scale systems for public employees and civil servants;
• It supplies the government and all interested parties with useful data for the analysis of wage rates of different groups of employees;
• International statistical data deliveries are provided from this database such as the October Inquiry of ILO and some Eurostat and OECD data requests;
• It is a valuable source of information for different research projects on national and international level;
• It provides useful information for potential Hungarian and foreign investors.

Differently from other countries, SES Hungary collects earnings data of the month of May. It is important to keep May as the reference month as it allows us to use the latest figures in the wage agreement process as early as November or December the same year. Since we kept May the reference month, the limitation remains that we can not collect the annual non-regular payments of the reference year, therefore, we collect the non-regular yearly earnings of the previous year. As described in the quality report of SES-2002 in great details this is not a serious setback, because the difference caused by this limitation is certainly less than the sampling error.

SES 2014 includes variables defined mostly by Eurostat (compulsory variables) applying classification systems that are harmonized with international classification systems. The Hungarian classifications are translated (FEOR/ISCO, Educational attainment/ISCED, TEÁOR/NACE) according to the instructions of HCSO. SES Hungary has one special variable: the ranking position of employees regulated by Labour Ministry Decree 6/1992. The position codes are not included in the database we send to Eurostat, therefore, we do not translate them either.

Classifications used:

NACE Rev 2.  - Business entities, budgetary and non-profit organisations are classified into different categories of NACE Rev 2. according to their main activities.

ISCO-08 – Employees are classified according to the International Standard Classification of Occupations '08.

ISCED11: Educational attainment of employees

NUTS - Updated territorial standards for statistical purposes are used.

Businesses with legal entities and other businesses employing at least 5 persons, all budgetary organizations, and selected non-profit institutions are covered by the survey, which are classified according to their main activities in NACE Rev. 2 classes and their size. Data by the occupation of employees is produced at 4 digit level by ISCO-08. Regional statistics are produced for NUTS1 and NUTS2 breakdown. The educational attainment of employees are published according to ISCED11..

All statistical concepts and definitions were harmonized with the requirements of Council Regulation 530/1999, the Commission Regulations 1738/2005 and 1916/2000.

In 2014 the National Labour Office collected earnings data of employees from business enterprises with legal entities, other businesses, budgetary organizations and selected non-profit institutions. The basic statistical unit of data collection was the local unit.

The Hungarian Business register's frame covers the whole statistical population.

The whole area of Hungary

SES data in Hungary are comparable since 2002 except for data by occupations (ISCO) and NACE classes. Comparable data for all sections according to the NACE Rev 2. classification are since 2008. Time series classified by NACE Rev. 1.1 are available from 2004 onwards until 2008.  Time series classified by ISCO_08 are available since 2011.

The reference month is May.

The first survey took place in 1992. SES harmonized surveys have been conducted since 2002. Time series classified by NACE Rev.2 are available since 2008. Time series classified by ISCO-08 are available since 2011.

-

Source of data: Structure of Earnings Survey

Base used for the sample: Business Register of HCSO

Sampling design:

Two different methods are used for the sampling:

• All employers over 50 employees are required to report about a sample of their employees: those who were born on specific dates in any month, in any year. We use three birthdates in the case of white-collar workers: the 5th, 15th and 25th day of any month in any year. Three times twelve days in a year results in a 36/365=0,0986 sampling ratio. For blue-collar workers within the business sector we use only two birthdates: the 5th and 15th of any month in any year, which results in a 24/365=0.0657 sample size. Within the budgetary sector we use three birthdates both for white and blue collar workers (the 5th , 15th and 25th). (One date means a 3,29% perfect probability sample, two dates 6,57% and three dates 9,86%)
• From those employers who have less than 50 employees a 20% random sample is chosen using the business register of the CSO, and those who were chosen have to report about each of their employees. The second method is based on a stratified random sample which is drawn by the CSO.
• In 2014 the CSO selected a 8% representative sample of micro enterprises (with more than 2 employees but less than 5 employees) as well which we included in the survey for the first time.

   For budgetary institutions, which report individually, the first sampling method is the same. However, most of these institutions are incorporated in a central payroll system and thus, for these institutions we get all the necessary data of each individual employee. In 2014 about 99,7% of public institutions reported this way.

  
  Retention/renewal of sampling units: Every year

   

  Sample size: 13,597 institutions from the government sector and 27,081 business units, altogether 40,678 units.

   

  Stratification:

  The surveys are based on stratified probability samples. The criteria for stratification are the following:

  - economic activity [in construction (two-digit level divisions 41–43) within the three-digit level group 42.1 the various four-digit level classes, otherwise the various groups; within divisions 45, 46, 47 and 56 groups 45.1, 46.9, 47.3 and 56.1, respectively, moreover, the other groups; within divisions 10, 18, 25, 31, 49, 52 and 85 classes 10.71, 18.12, 25.11, 31.09, 49.41, 52.29 and 85.59, respectively, furthermore the other classes],

  - size (categories: 20–49, 10–19 and 5–9 persons employed),

  - place of the headquarter (categories: Budapest and the countryside).

  The total samples sizes are determined by taking into account both accuracy demands and cost limitation. Grossing up by simple inflation, the sample sizes for the various strata are determined by

   

   

  where Nj is the size of the stratum, u = 1.96 is the standard normal percentile belonging to the level of confidence 0.95, Cj is the coefficient of variation calculated from data of previous surveys and vj is the relative margin of error (the radius of the relative confidence interval) which we would like not to exceed. Then the percentage sampling rate is given by the formula

   

  The values of vj are determined in such a way that the sample scheme should correspond to a modified Neyman allocation.

  Data of those non-respondents of which – on the basis of information on the cause of non-response – we suppose that they would have sent us a questionnaire with only zero values are imputed by zero. Data of the other non-respondents are imputed by the mean value of the amalgamated Budapest–countryside stratum. Though the variables under consideration do not have very high variance between enterprises, we pay attention to outliers. In order to compare enterprises of different strata, we standardize the variables and modify them according to the (sampling) size of the stratum. (This modification is necessary because the fewer is the data the higher is the probability of higher value.) Enterprises with modified value larger than a threshold are considered to be outliers. (The threshold values for the various variables are determined on the basis of mathematical and experimental considerations.) Outliers are not included in the gross up procedure, they are handled separately.

  In the course of the processing indicators are estimated by strata for the sampled enterprises and the enterprises enumerated completely. In case of the non-specific characteristics (occupied jobs, and job vacancies), the population total is estimated directly from the elementary data. Within the various sampled strata the sampling weight

   

  is determined, and the monthly (last month of the quarter) population total Yj is estimated by the Horwitz-Thompson estimator with equal selection probabilities:

  
   

  where yj is the sample total in the jth stratum.

   

The totals estimated in the above way are broken down first among the four-digit level economic activity classes, then among the counties of the countryside.

The survey is conducted once a year. The reference month of the data is May.

The deadline for data collection for business enterprises and non-profit organizations was June 27, 2014 and for budgetary institutions July 11, 2014. Data suppliers could send in their statistics electronically through the data recording system of the National Employment Office, by email to tarifa@lab.hu and on paper.  35 % of respondents sent their data electronically, 16% by email and 49% on paper.

The validation process is carried out in a planned way every year. The first step of the revision process includes the correction of simple coding errors and the inputation of missing codes. The simple errors include the following:

• missing variables: sometimes missing data are collected from the respondents personally, in other cases they are imputed
• slipped fields: they can be detected through the checking process and in most cases can be corrected
• duplicate reports or records: duplicate reports are excluded
• outstanding values in any field: checked and corrected when necessary
• incorrect code values: checked and corrected

  In addition, we identify data that are in contradiction with each other through logic tests. We compare different fields within the individual records, like education, occupation, educational attainment and wage and correct these errors by applying predefined algorithms based on regular relations among data.

The validation process was completed from September 1 through November 30, 2014. Data were tested before data recording to find simple coding errors and duplicates, and after recording several logic tests were applied to identify incorrect coding, missing data and data errors. Recontacts were made when it was necessary. The most problems were revealed in the case of data sent through e-mail. Electronically processed data had the least errors because the program has built in checking algorithms that signal errors during data upload. Erroneous data were corrected through a thorough revision process.

After the validation process, sampling weights are determined for the data.

In the case of business units employing less than 50 employees, where data suppliers are subject to representative sampling by HCSO, we apply a two-step weighting procedure. First, data are extrapolated by a weight figure defined by dividing the number of all employees by the number of employees included on the data sheet of a given organization. (As a rule, businesses with less than 50 employees have to provide data on every employee. However, they have to exclude those who were absent for more than three days in the reference month not to distort earnings data.)  In the second step estimations are made by assigning a second weight to each individual. The second correction weight is the reciprocal of the sampling ratio where the sampling ratio is the ratio of the total number of employees in a given size band provided by HCSO and the number of employees we received data for in that size band. If, for example, on the basis of a 5% sampling ratio, we assign a multiplier of 1:0.05=20, this means that he/she represents 20 members of the total population and similarly a sampling ratio of 8.25% implies a weight of 12.

Business organizations with more than 50 employees and a small part of government institutions provide data only on a selected part of their employees. In their case the weight figures are defined by dividing the number of all full time employees of the organization by the number of employees included in the sample. The same procedure is applied to the government sector data we receive from the central state payroll system.

This means that data included in the tables are the results of statistical corrections. Values pertain to the total represented population and as such should be regarded as estimates about the total represented population. The estimates are more accurate and reliable if the sub-population, the calculations are based on, are larger. It is advisable to treat breakdowns very cautiously and if reliability drops below a certain predefined level, data should be marked or excluded from publication.

Monthly earnings are adjusted to reflect average yearly earnings using total non-regular bonuses and other non-regular income divided by 12 of the previous year.

We do not revise data retrospectively.

–

We do not have quality objectives defined by a decree or a policy.

In the past couple of years the National Employment Office introduced changes in the data collection and validation process that aimed at improving data quality. We developed the data recording web page by working out and incorporating new logic tests in the system and simplifying its data upload function. More and more data have been sent electronically and the program is increasingly user friendly.

We have maintained close contacts with payroll software developer companies and consulted with them on systematic errors in the delivered survey data to filter out possible programming errors. Continuous communication has improved data quality as well.

During the data collection process we provided assistance to data suppliers. We answered 100-200 calls a day. 

Errors due to the sampling error have been measured by the relative standard error. If the standard error of a cell is between 10% and 20%, the number is marked by *. However if the standard error is larger than 20%, we do not publish the data.

-

We do not analyse user satisfaction through a specific survey, however we ask for feedback concerning our data from our users.

In most cases we can satisfy specific requests as well, the only occasional limits are the rules of data protection.

SES-2014 is complete and contains all compulsory variables required by Commission Regulation 1738/2005. All the necessary breakdowns were provided (NACE, ISCO, ISCED, NUTS), the transcoding was made on the required levels (NACE Rev. 2 classes, ISCO-08 first digit, ISCED 0-6 and NUTS level 1).  This means that SES-2014 Hungary is completely comparable on EU level. The only difference is that we use May as our reference month instead of October, however, this doesn’t cause serious problems.

Our survey is conducted every year, even though EU regulations require the completion of the survey only every four years. All surveys are conducted according to EU regulations.

We do not compute a data completeness rate.

-

Overall accuracy is not measured.

The survey was partly conducted through a stratified probability sample provided by the Business Register of the HCSO. The sampling units were the business enterprises selected from the relevant NACE classes and size groups of the Business Register. Due to the sampling procedure data may be subject to random deviation, that is sampling error.

Due to the probability sampling method, there is no systematic bias in the estimations.

The sampling error is measured by the standard error or the relative standard error. If the standard error of a cell is between 10% and 20%, the number is marked by *. However if the standard error is larger than 20%, we do not publish the data.

The coefficients of variation for monthly and hourly earnings of full-time and part-time employees, according to the breakdowns required were computed and presented in Attachments C and D. The CVs are high because the dispersion of wages is much bigger among individuals than in the case of company averages.

–

The Business Register of HCSO has direct online connection with the registry court in Hungary. The register is updated monthly. However, possible under- or over coverage may occur because the birth, merger and demerger of business units are difficult to follow day by day. The set up of new companies can be tracked easily because new companies have to request for an identification number. However, data tracking is much more difficult in the case of company closings. To avoid under- or over coverage to the possible rate, we use the latest updated version of the sampling frame.

Under- or over-coverage may occur because of misclassification of the number of employees by size bands. In most cases this does not cause any problems, because for the grossing up of data we use the actual total number of employees, given on the cover-sheet of the report by each employer.

Misclassification can also occur if the main activity (NACE code) of the local unit is incorrect.

Multiple listing is not known.

We do not compute an over-coverage rate.

Multiple listing is not known.

In our survey measurement errors are reporting errors. The most important sources of these include:

• Erroneous coding of the firm identification number or the activity code;
• Data entry errors by respondents;
• Data entry errors during the recording process;
• Possible errors during data transmission or transformation;
• Possible errors during data processing at NLO;
• Possible errors during the transformation of national codes of education into ISCED codes.

 A special feature of the survey in Hungary is that the majority of data in the budgetary sector come from a central payroll system. It means that theoretically measurement errors are not possible, unless some variables are missing from the central system or there are errors in the system. The central payroll system uses the same terms and definitions that are determined and used by the Central Statistical Office for statistical purposes.

The survey is part of the annual National Programme of (obligatory) Statistical Reports. It is compulsory for all firms over 50 employees, for all budgetary institutions and for firms in the business sector with less than 50 employees which were selected for the sample. In spite of obligatory reporting, however, a number of business units including some larger companies do not respond to the survey.

In 2014 the response rate in the business sector was 64% in case of organizations with 5 employees or more. In case of smaller firms, especially of those with less than 20 employees, the response rate was lower (around 55%) than in the case of larger business units, where the response rate was between 80-99%. However, the non-response of smaller units can partly be corrected by using the grossing up method described in 3.5. In case of smaller companies where a 20% sample was selected from the complete list of companies, the grossing up factors are determined for the individual size bands by using the total number of employees in the appropriate size band according to the HCSO’s quarterly institutional survey.

Using arrival lists we try to collect data from the most important respondents by calling them on the phone. In most cases we succeed, however, a few remain missing from the survey. The lack of our capacity limits these efforts.

In the budgetary sector the response rate is excellent, in 2014 it reached 99.9% thanks to the well functioning central payroll system.

The non-response rate causes biases concerning the total number of employees and their distribution by size bands, but these biases are small concerning the estimates of monthly or annual earnings and for hours paid, which are the most important results of the survey.

A general item non-response rate is not calculated. Although item non-response rate was low in the case of most variables in 2014, it was much higher than the average related to two variables. Educational attainment and the job position variable was missing in 8.5% of the cases.  Missing values were inputed after contacting data suppliers again or on the basis of predefined algorithms.

The number of errors corrected are not calculated.

We do not calculate an imputation rate.

This is not applicable in our case, because we don’t use model assumptions for the Structure of Earnings Survey.

We do not apply seasonal adjustment.

We do not have planned revisions for our data, generally we do not correct our data retrospectively. This is also due to the fact that SES data serve as a basis for official earnings statistics in court procedures. Retrospective modifications in data may lead to confusions in court procedures

Right after data compilation, before the main results of the database are published, the data are compared to HCSO institutional statistics. If major differences occur, we revise the database.

We did not have data revision in 2014.

–

• Respondents were informed about the survey in November of the previous year after the National Statistical Council accepted the National Data Collection Program.

  The reference month of the Hungarian survey is the month of May. Key data collection dates in 2014 were as follows:

  • The questionnaires and instructions were sent out to the respondents at the beginning of May 2014
  • The deadlines for sending back the questionnaires:

  - for companies with less than 300 employees:  27^th June 2014

  - for companies over 300 employees:                 11^th July 2014

  - for all budgetary institutions:                           11^th July 2014

   

  • Post collection phases:

     

  - receiving the reports, manual checks, recalling missing responses by using computerized arrival lists: continues until the end of August

  - data entry and compiling the reports received electronically: parallel with the previous phase

  - logical checks, data corrections: September – October 2010

  - data entry for late arrivals: September 2014

  - data validation: September - November 2014

  - data compilation: December 2014

   

  • Preliminary results given to the social partners to use in wage negotiations: December 2014
  • Final results published: February 2015 (on CD)
  • Transforming the database of SES 2014 according to the requirements and sending it to the Eurostat through the HCSO: October 2015
  • Sending the final version of the database to Eurostat through the CSO: July 2016
  • Data processing for the quality report: December 2016
  Quality report:  December 2016

First and final results were published in February 2015. The time lag between the first result and the reference period was +8 months.

First and final results were published in February 2015. The time lag between the final result and the reference period was +8 months.

Deadlines for data delivery and publication were all kept in 2014.

Deadlines for data delivery and publication were all kept in 2014.

The SES survey is collected from a range of sources in Hungary. We are taking over and compiling data produced for statistical and not for administrative purposes. Therefore, the data are coherent. The national concepts, terms and definitions are harmonized with the European ones. The classification by sectors corresponds with the NACE coding system generally used in the EU. The Hungarian classification of occupations (FEOR-93) is similar to ISCO-88. The structure and principles are the same, but they are equivalent only on the 1^st digit level. Transcoding is possible on the two and three-digit levels but not possible on the 4-digit level. There is no generally used classification of educational attainment in Hungary, but the level of education may be transcoded to ISCED (0 to 6). The statistical unit of the survey is the local unit. The industrial classification comes from the register of employers of the Central Statistical Office.

The definition of the variables and classifications of the SES survey in Hungary are harmonized with the requirements of EU regulations.  Therefore, Hungarian SES data are comparable on the EU level.

Not calculated

SES data in Hungary are comparable since 2002 except for data by occupations and NACE classes. The Hungarian survey was harmonized with EU regulations in 2002. The scope of the survey was extended, new variables were introduced, however, the definitions of old variables remained the same. When the ISCO and NACE classifications changed, we did not review the data retrospectively and did not publish data according to the new classification systems.

The structure of the survey has not changed since 2002, therefore comparability is assured since then for most of the breakdowns (such as regional, gender, age, ownership). However, as new NACE codes were introduced in 2008 comparisons between SES-2006 and SES-2014 and SES-20010 and SES-2014 data broken down by industry is not possible. In 2011 new ISCO codes were introduced, therefore SES-2014 data by occupations can not be compared with SES-2006 and SES-2010 data.

The comparability of our results with other statistics is limited. However, the main outcomes of the institutional survey data of HCSO are comparable with SES data. Before publishing our final results, we compare the aggragated results of the two surveys every year.

–

According to Commission Regulation 698/2006 we investigated the coherence regarding the compensation per employees between SES and SNA-based figures. The results can be found in the table below.

The table shows that in almost all NACE sections the SNA based compensation per employee figures are higher than the average earnings data from the Structure of Earnings Survey. The possible

• The difference in coverage: SNA data are derived from the balance sheets of all companies while SES collects individual earnings data.
• National Account data include also estimates on the grey and black economy, which can be significant in certain sectors of the national economy.

Not relevant.

–

Not relevant, we do not publish a press release.

We do not publish data on paper since 2014, however a predefined set of tables are accessible on the internet:

a) For the open public:

• some of the most important results are accessible for everybody, free of charge, on the website of the National Labour Office (www.munka.hu), in English as well
• some of the aggregated results of the previous year are sent out to the respondents of the new survey  with the questionnaires

   b) For the interested ministries, trade unions and employers organizations:

• the results of model computations to prepare negotiations on minimum wage decisions and macro level wage agreements are also computed

   c) For selected important users (social partners, Directorates of CSO, Regional Labour Centres, Universities, Research Institutes, ministries, libraries) a four-volume publication was published every year on basic wages and earnings (cca 4*200 pages) and since 2014 data are only accessible on the internet:

• about the Hungarian economy as a whole
• about the business and the non-profit sectors
• about the government sector

by regions (counties, NUTS 3 level)

Microdata are not accessible online.

We offer consultations for data users if required. We do not report on the number of consultations

Users of the whole database:

• Each year the CSO takes over the whole database to fulfill important international reporting obligations (ILO, Eurostat, OECD)
• Hungarian and international research institutes take over the whole database for research purposes (without the individual identifiers of the employers)

Certain government agencies also take over the whole database (such as the Hungarian National Bank).

Trade unions, companies, potential investors etc. may order for special data processing for reimbursement costs taken into consideration the rules of data protection.

Methodological comments and explanations are attached to the data and the tables we disseminate or publish.

There is no special quality management related to documentation, however, documents are continuously updated.

Not calculated.

We offer consultations for data users if required. We do not report on the number of consultations.

The total costs associated with the collection and production of the survey amounted to HUF 29,200,000 that is EUR 94,500 in 2014.

-

We do not have officially approved policy on confidentiality.

The following legal regulations ensure confidentiality:

-       Law no. 46/1993 on „Statistics” and Law no. 155/2009 on „Data protection” and Law no. 112/2011 on „Information self-determination and information freedom.”

• Director General’s Order no. 15-03/F/2013 on „Data protection and data security” at the National Labour Office. The Order is reviewed annually and includes regulations related to data management, data processing and confidentiality.

Principles concerning data protection of the SES survey include the following :

• Identifiable data on institutions can be transferred only with the written consent of the leader of the relevant institution.
• We do not publish data if the number of data suppliers in the relevant cell is less than 4.
• When we deliver the database with micro data for research, we enter into a contract with the recipient institutions. In the contract we specify that data can not be transferred to a third party.
• We transfer the microdata with a special, anonymized identifier which substitutes for the real identifiers. If the data recipient would like to connect data from more than one year then the anonymized identifier will contain of the same numbers each year.
• Researchers can connect SES data with other databases only with the permission of the National Information Authority. In these cases we anonymize data with a special “hash” coding program.
• The hardware and software background of the survey follows the instructions of data protection of the National Labour Office. The database is stored on a separate, designated computer, which is managed by the system manager of the SES survey. The Labour office protects the data with a strong firewall.
• The person in charge of data protection is the system manager of SES.

The required results for tabular analyses are attached in Attachments A and B.  The distributions of full-time (FT) employees and part-time (PT) employees are provided separately according to the required breakdowns. The results of SES-2010 can be compared to the results of SES-2006, which were attached to the Quality Report of SES-2006, except for the breakdowns by NACE due to the changes in the NACE coding system in 2008.